Skip to content

mtmd: Unlimited-OCR fix max_tiles, setting in converter#25614

Open
sfallah wants to merge 1 commit into
ggml-org:masterfrom
sfallah:sf/dsocr-preproc-max-tiles
Open

mtmd: Unlimited-OCR fix max_tiles, setting in converter#25614
sfallah wants to merge 1 commit into
ggml-org:masterfrom
sfallah:sf/dsocr-preproc-max-tiles

Conversation

@sfallah

@sfallah sfallah commented Jul 13, 2026

Copy link
Copy Markdown
Contributor

Description

Unlimited-OCR is sharing DeepSeek-OCR v1's projector type, so it inherited v1's max_tiles=9, while the actual HF default is 32.
The max_tiles now comes from the GGUF params, with the old values as fallback.

Requirements

AI usage disclosure: YES - I used AI assistance for code review, debugging, implementation checks, and testing. I have reviewed the submitted changes and take responsibility for the full contents of this PR.

@sfallah sfallah requested review from a team and CISC as code owners July 13, 2026 07:32
@github-actions github-actions Bot added mtmd Related to multimodal functionality (video/image/audio) conversion labels Jul 13, 2026
@sfallah sfallah force-pushed the sf/dsocr-preproc-max-tiles branch 2 times, most recently from 36ca480 to 8f2a0c8 Compare July 13, 2026 11:27
@sfallah sfallah force-pushed the sf/dsocr-preproc-max-tiles branch from 8f2a0c8 to 98dfe4a Compare July 13, 2026 14:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

conversion mtmd Related to multimodal functionality (video/image/audio)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant